Personalized Video Recommendation Integrating User Portrait and Collaborative Filtering

Abstract

In order to improve the quality of recommendation and user-perceived service, this paper constructs a personalized recommendation model that integrates user portraits and collaborative filtering. First, build a portrait label system based on user characteristics, use time decay function and TF-IDF to obtain label weights, calculate user similarity through user feature labels, and merge it with user similarity obtained by user-based collaborative filtering algorithm to reconcile the weights. Obtain the comprehensive similarity of users, then take Top-N in descending order to form the final personalized recommendation. This paper conducts experimental verification through Douban website, and uses offline experiments to prove that compared with a single algorithm, a video personalized recommendation model that combines user portraits and collaborative filtering algorithms can improve the quality of personalized recommendations to a certain extent.